Francis et al.; Quantifying annual variation in catchability 



299 



in the same years). These patterns would be very unhkely 

 to occur by chance alone if there were no between-species 

 correlations. For series 5, 14 of the 22 species had their two 

 highest biomass indices in the same years ( 1994 and 1996). 

 The probability that an outcome as extreme as this would 

 occur by chance alone (assuming no correlations) is only 

 6.4 X 10-6. poj. series 6, the probability is 9.8 x 10* (here 

 17 of 25 species had their two highest years in 1996 and 

 1998). These very low probabilities are clear evidence that 

 there are sometimes strong between-species correlations in 

 a survey series. We will argue below that the main cause 

 of these correlations is that catchability was extreme (for 

 many species) in some years. 



In our more formal analysis, 16 of a total of 94 survey 

 years were found to be significantly extreme (nine with 

 high catchability, and seven with low), and there were 

 eight survey series for which no years were extreme (Fig. 5). 

 (Note that the vertical distance between the broken lines in 



3.0- 



25- 



2.0- 



1.5- 



1.0- -<, 



0.5- 



Series 5 



» O.Q-^ 



1993 



1994 



1995 



1996 



2.0- 



1.5- 



10- 



0.5- 



O.Q- 



Series 6 



1997 



1998 



1999 



2000 



Year 



Figure 4 



Standardized biomass indices, plotted against year, for 

 trawl survey series 5 and 6. For each species in a series, the 

 biomass indices were standardized by dividing by the mean 

 index for that species. 



Figure 5, which indicates how extreme a mean rank needs 

 to be to be judged significant, decreases with increasing 

 number of species and with decreasing number of survey 

 years. ) We also investigated three modifications to the above 

 procedure for identifying extreme years to see whether they 

 might be useful. None was (see Appendix for details). 



For some extreme years the biomass ratio statistics 

 (Table 5) are so large that it is unlikely that actual bio- 



